A Multi-criteria approach for the evaluation of rule interestingness

This paper studies several criteria for evaluating rule interestingness. It first reviews some rule interestingness principles with respect to the widely-used criteria of coverage, completeness and confidence factor of a rule. It then considers several additional factors (or criteria) influencing rule interestingness that have been somewhat neglected in the literature on rule interestingness. As a result, this paper argues that rule interestingness measures should be extended to take into account the additional rule-quality factors of disjunct size, imbalance of the class distribution, attribute interestingness, misclassification costs and the asymmetry of classification rules. The paper also presents a case study on how a popular rule interestingness measure can be extended to take into account the proposed additional rule-quality factors.